Multiscale temporal and spatiotemporal analysis of wheat blast infection-like dynamics using vertical plant stratification, regression and Markov chain approaches

Abstract

The spatiotemporal progress of wheat blast (WB) epidemics within the plant canopy remains poorly known due to complex pathogen–host–environment interactions. Although deterministic methods are popular and useful, robust stochastic methods, such as generalized additive models for location, scale and shape (GAMLSS) and probability matrix or Markov transition model (MTM), have seldom been used to analyse plant disease epidemics. Hence, both methods were employed to derive valuable insights into WB epidemiology at the vertical canopy level. We conducted experiments in three climatic zones in Bolivia, using three wheat cultivars, with disease data corresponding to different canopy positions (lower, L; middle, M; flag leaf, F; and spike, S). Using WB severity data (AUDPC, progress rate and maximum severity [Kmax]), we implemented the GAMLSS and MTM to test our hypothesis that WB is affected by host resistance, location and canopy level. Results showed that the AUDPC, progress rate and Kmax differed across sites, cultivars and canopy positions. Clearly, L and M canopies showed a lower progress rate than F and S. The disease showed an ascending movement from L and M canopies to F and S across locations and cultivars. However, descending transitions also occurred from M to L early or F to M canopy later in the season. Both ascending and descending movements can arise at a single state or several recurrent states, indicating indirect evidence of autoinfection within the canopy before and after spike emergence. Our findings contribute knowledge to improve monitoring and managing WB.